A disordered acidic domain in GPIHBP1 harboring asulfated tyrosine regulates lipoprotein lipaseKristian K. Kristensena,b, Søren Roi Midtgaardc, Simon Myslinga,b,d, Oleg Kovrove, Lars Bo Hansenf,Nicholas Skar-Gislingec, Anne P. Beigneuxg, Birthe B. Kragelundh, Gunilla Olivecronae, Stephen G. Youngg,i,1,Thomas J. D. Jørgensend, Loren G. Fongg, and Michael Plouga,b,1

aFinsen Laboratory, Rigshospitalet, DK–2200 Copenhagen N, Denmark; bBiotech Research and Innovation Centre, University of Copenhagen, DK–2200Copenhagen N, Denmark; cStructural Biophysics, Niels Bohr Institute, University of Copenhagen, DK–2100 Copenhagen Ø, Denmark; dDepartment ofBiochemistry and Molecular Biology, University of Southern Denmark, DK–5230 Odense M, Denmark; eDepartment of Medical Biosciences, Umeå University,SE-901 87 Umeå, Sweden; fZealand Pharma, DK–2600 Glostrup, Denmark; gDepartment of Medicine, University of California, Los Angeles, CA 90095;hDepartment of Biology, University of Copenhagen, DK-2200 Copenhagen N, Denmark; and iDepartment of Human Genetics, University of California, LosAngeles, CA 90095

Contributed by Stephen G. Young, May 17, 2018 (sent for review April 20, 2018; reviewed by Jay D. Horton and Sampath Parthasarathy)

The intravascular processing of triglyceride-rich lipoproteins de-pends on lipoprotein lipase (LPL) and GPIHBP1, a membrane proteinof endothelial cells that binds LPL within the subendothelial spacesand shuttles it to the capillary lumen. In the absence of GPIHBP1, LPLremains mislocalized within the subendothelial spaces, causingsevere hypertriglyceridemia (chylomicronemia). The N-terminal do-main of GPIHBP1, an intrinsically disordered region (IDR) rich inacidic residues, is important for stabilizing LPL’s catalytic domainagainst spontaneous and ANGPTL4-catalyzed unfolding. Here, wedefine several important properties of GPIHBP1’s IDR. First, a con-served tyrosine in the middle of the IDR is posttranslationally mod-ified by O-sulfation; this modification increases both the affinity ofGPIHBP1–LPL interactions and the ability of GPIHBP1 to protect LPLagainst ANGPTL4-catalyzed unfolding. Second, the acidic IDR ofGPIHBP1 increases the probability of a GPIHBP1–LPL encounter viaelectrostatic steering, increasing the association rate constant (kon)for LPL binding by >250-fold. Third, we show that LPL accumulatesnear capillary endothelial cells even in the absence of GPIHBP1. Inwild-type mice, we expect that the accumulation of LPL in close prox-imity to capillaries would increase interactions with GPIHBP1. Fourth,we found that GPIHBP1’s IDR is not a key factor in the pathogenicityof chylomicronemia in patients with the GPIHBP1 autoimmune syn-drome. Finally, based on biophysical studies, we propose that thenegatively charged IDR of GPIHBP1 traverses a vast space, facilitatingcapture of LPL by capillary endothelial cells and simultaneously con-tributing to GPIHBP1’s ability to preserve LPL structure and activity.

hypertriglyceridemia | electrostatic steering | intrinsically disorderedregion | intravascular lipolysis | autoimmune disease

Lipoprotein lipase (LPL) is the central and rate-limiting en-zyme for the intravascular lipolytic processing of triglyceride-

rich lipoproteins (TRLs). The intravascular hydrolysis of tri-glycerides releases lipid nutrients for vital tissues (e.g., heart,skeletal muscle, adipose tissue) (1–3). Parenchymal cells (e.g.,myocytes and adipocytes) synthesize LPL and secrete it into theinterstitial spaces, but LPL’s site of action is within the capillarylumen. The secreted LPL is retained in the interstitium, where itforms a dynamic reservoir via transient interactions with heparansulfate proteoglycans (HSPGs). A membrane protein of capillaryendothelial cells, glycosylphosphatidylinositol-anchored high-density lipoprotein binding protein 1 (GPIHBP1), capturesLPL from the interstitial spaces and shuttles it across endothelialcells to the capillary lumen (4–6). Having reached the capillarylumen, the LPL–GPIHBP1 complex is the key functional unit forTRL processing, mediating both the docking of circulating TRLs tocapillary endothelial cells (5) and the rapid hydrolysis of theirtriglyceride content (7). The resistance of LPL–GPIHBP1 com-plexes to inactivation by physiologic inhibitors [e.g., angiopoietin-like (ANGPTL) proteins 3, 4, and 8] serves to focus catalyticallyactive LPL along the capillary lumen (8–13).

A wealth of genetic and experimental evidence highlights theimportance of LPL and GPIHBP1 in maintaining normal plasmatriglyceride levels. Homozygous or compound heterozygous loss-of-function mutations in LPL or GPIHBP1 cause severe hyper-triglyceridemia (familial chylomicronemia) (1, 2). This syndrome isassociated with life-threatening bouts of acute pancreatitis (1). Also,mice with a deletion of Gpihbp1 (Gpihbp1−/−) or mice harboring aGpihbp1 missense mutation known to cause disease in humans(Gpihbp1C63Y/C63Y) develop severe chylomicronemia (4, 14). In thesemouse models, LPL remains confined within the interstitial spaces,never reaching the capillary lumen. Recent studies have shown thatautoantibodies against GPIHBP1 cause some cases of acquired chy-lomicronemia (GPIHBP1 autoantibody syndrome) (15). GPIHBP1autoantibodies bind to GPIHBP1 and block LPL binding, therebyabolishing the formation of the key functional unit (GPIHBP1–LPL)required for triglyceride hydrolysis within capillaries (15, 16).To orchestrate the assembly of this functional unit, GPIHBP1

evolved unique structural properties (2). Despite being a rela-tively small protein (131 residues), GPIHBP1 is highly asym-metrical with an N-terminal intrinsically disordered region (IDR)rich in acidic residues, a disulfide-rich core Ly6/uPAR (LU)domain, and a short C-terminal region that tethers the protein tothe cell membrane via a glycolipid anchor (2, 17). This multifaceted

Significance

Dietary fats absorbed by intestinal enterocytes are packaged intochylomicrons, which circulate in the bloodstream until dockingalong capillary endothelial cells, primarily in striated muscle andadipose tissue. There, a functional unit composed of lipoproteinlipase (LPL) and GPIHBP1 hydrolyzes the triglycerides in chylomi-crons, releasing free fatty acids for use by surrounding paren-chymal cells. The current study describes the interactions betweenLPL and GPIHBP1 and describes how a disordered region ofGPIHBP1 results in order and functional stability within LPL. Theelucidation of LPL–GPIHBP1 interactions sheds light on the regu-lation of intravascular triglyceride processing and provides in-sights into plasma triglyceride metabolism in health and disease.

Author contributions: K.K.K., S.G.Y., and M.P. designed research; K.K.K., S.M., O.K., L.G.F.,and M.P. performed research; L.B.H., A.P.B., and G.O. contributed new reagents/analytictools; K.K.K., S.R.M., S.M., O.K., N.S.-G., B.B.K., G.O., S.G.Y., T.J.D.J., L.G.F., and M.P. ana-lyzed data; and K.K.K., S.G.Y., and M.P. wrote the paper.

Reviewers: J.D.H., University of Texas Southwestern Medical Center; and S.P., University ofCentral Florida.

Conflict of interest statement: S.G.Y. and Jay D. Horton were both principal investigatorson a consortium grant. They did not collaborate directly.

Published under the PNAS license.1To whom correspondence may be addressed. Email: sgyoung@mednet.ucla.edu orm-ploug@finsenlab.dk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1806774115/-/DCSupplemental.

Published online June 13, 2018.

E6020–E6029 | PNAS | vol. 115 | no. 26 www.pnas.org/cgi/doi/10.1073/pnas.1806774115

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

structure plays different roles in the regulation of LPL activity inthe capillary lumen. The folded LU domain is responsible for theformation of a stable complex with LPL, while the IDR stabilizesthe hydrolase domain of LPL against spontaneous unfolding (12,17). Nonetheless, only full-length GPIHBP1—and not the LUdomain or the N-terminal IDR alone—efficiently protects LPLagainst ANGPTL4-catalyzed unfolding (12).The concept that IDRs can regulate cellular functions via spe-

cific interactions emerged from studying intracellular proteinsinvolved in cell signaling and nucleic acid binding (18). In thissetting, specific phosphorylations in IDR sequences altered pro-tein dynamics, adding a regulatory switch to IDR function (19, 20).The current study brings this concept—that IDRs can mediatespecific protein interactions—into the realm of intravascular tri-glyceride metabolism (12, 17). First, we demonstrate that a con-served tyrosine within GPIHBP1’s acidic IDR undergoes O-sulfation and that this modification affects LPL interactions andLPL stability. Second, we investigate the importance of the N-terminal acidic IDR for LPL binding, for stabilizing LPL activ-ity, and for capturing LPL from the pool of HSPG-associated LPLwithin the subendothelial spaces. Finally, we assess the relevanceof GPIHBP1’s acidic IDR to the GPIHBP1 autoantibody syn-drome, in which autoantibodies against GPIHBP1 block LPLbinding and lead to severe chylomicronemia (15).

ResultsPosttranslational Modifications of Human GPIHBP1. A secreted ver-sion of human GPIHBP11–131 was purified from the medium oftransfected Drosophila S2 cells (17). The molecular mass of the

principal protein species was 15,722.6 Da as determined by massspectrometry (Fig. 1A, Inset). Posttranslational modificationsaccounted for 1,119 Da. N-glycanase treatment revealed that theGPIHBP1 contained one N-linked glycan (Fig. 1A), a typicalbiantennary glycan (21, 22) attached to Asn58 (SI Appendix, Fig. S1).

GPIHBP1 Carries a Tyrosyl-O-Sulfate Modification. After subtract-ing the mass of the N-linked glycan (1,038.96 Da), the mass ofGPIHBP1 remained 80 Da greater than expected. This differenceis compatible with one phosphorylation or sulfation. Only a minorfraction of GPIHBP1 lacked the 80-Da modification (Fig. 1A, In-set). The only posttranslational modification in a truncated humanGPIHBP1 lacking the N-terminal acidic IDR (GPIHBP134–131)was the N-linked glycan (Table 1), strongly suggesting that the80-Da modification is located within the first 33 residues ofGPIHBP1. We verified that assignment by showing that themodification was present in a GPIHBP11–33 peptide released fromfull-length human GPIHBP1 with trypsin (Table 1). Peptide pro-filing narrowed the possible modification sites to residues 10–21,and Tyr18 was the only hydroxy-amino acid in that segment. Todetermine if Tyr18 was indeed the modification site, we created aGPIHBP1 mutant in which Tyr18 was replaced by phenylalanine(GPIHBP11–131/Y18F). This single amino acid substitution elimi-nated the 80-Da modification in GPIHBP1 (Table 1).Several lines of evidence showed that Tyr18 in GPIHBP1 carries

a sulfate rather than a phosphate. First, phosphorylated, but notsulfated, versions of GPIHBP11–33 peptides were sensitive tophosphatase (SI Appendix, Fig. S2 A–D), and GPIHBP11–33 thathad been released from GPIHBP11–131 with trypsin was resistant

Fig. 1. Posttranslational modifications of recombi-nant human GPIHBP1. (A) Human GPIHBP11–131 pro-duced in Drosophila S2 cells was analyzed by SDS/PAGEand Coomassie blue staining (nonreduced, lane 1; re-duced and alkylated, lane 2). Mass spectra are shownfor intact GPIHBP1 before (Inset) and after N-glycanasetreatment under native conditions, which reduces themolecular mass by 1,038.5 Da corresponding to onepaucimannosidic N-glycan with a core fucose(Man3GlcNAc2Fuc), the archetypical insect cell glycan(21). The asterisks indicate the loss of 17 Da due tothe formation of pyroglutamate at the N-terminalglutamine in human GPIHBP1. (B) Sensorgrams show-ing the interactions between immobilized anti-sulfotyrosine mAb 1C-A2 (1,280 RU) and GPIHBP11−131

(blue curve), GPIHBP11−131/Y18F (green curve), and buffer(black curve). The SPR sensorgrams were recorded witha BiacoreT200 instrument and are shown for sequentialinjections of twofold dilutions of GPIHBP1 from 125 nMto 2 μMwithout intervening regenerations. The affinitybetween GPIHBP11−131 and mAb 1C-A2 was 5.4 μM(determined from several runs up to 16 μM GPIHBP1;see Inset).

Kristensen et al. PNAS | vol. 115 | no. 26 | E6021

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

to phosphatase (SI Appendix, Fig. S2 E and F). Second, a mono-clonal anti–Tyr-OSO3 antibody (1C-A2) (23) bound to purifiedGPIHBP11–131 but not to GPIHBP11–131/Y18F (Fig. 1B). Third,Sulfinator (24) and GPS-TSP (25) programs identified Tyr18 as alikely sulfation site (SI Appendix, Fig. S3).Alignments of GPIHBP1 sequences from different phylogenetic

orders of Mammalia revealed an N-terminal IDR that was in-variably present and enriched in acidic residues, although the aminoacid sequence varied considerably (SI Appendix, Fig. S3). Thetyrosyl-O-sulfation site (i.e., Tyr18 in mature human GPIHBP1) waslargely invariant, with only a few exceptions (e.g., members of theEquidae family). To determine if the conserved tyrosine was sul-fated in the GPIHBP1 of other mammalian species, we expressedmouse GPIHBP1, excised the acidic peptide with trypsin, and thenconfirmed by mass spectrometry that the acidic peptide was sul-fated (SI Appendix, Fig. S2 G and H).

Stoichiometry and Affinity of LPL Binding to GPIHBP1. To define thestoichiometry of the LPL–GPIHBP1 interaction, we incubated3 μMLPL with increasing concentrations of intact GPIHBP11–131 orthe disordered polypeptide GPIHBP11–33 (at concentrations abovethe Kd). We chose native PAGE to assess LPL–GPIHBP1 complexformation because of the large differences in pI values and molec-ular masses of the two proteins (9.5 and 50.5 kDa, respectively, forLPL vs. 4.1 and 15.7 kDa, respectively, for GPIHBP1). The titrationresulted in the formation of an LPL–GPIHBP1 complex with adistinct electrophoretic mobility (Fig. 2). Scanning of Coomassieblue-stained bands revealed that the stoichiometry of the LPL–GPIHBP1 complex was 1:1, regardless of whether the titra-tions involved GPIHBP11–131 (Fig. 2A), GPIHBP11–33 (Fig.2B), or GPIHBP11–131/Y18F (Fig. 2C).To assess the impact of tyrosyl-O-sulfation on the kinetics of LPL

binding, we measured the association (kon) and dissociation (koff)rate constants between GPIHBP1 in solution and LPL captured onthe LPL-specific monoclonal antibody 5D2. Using surface plasmonresonance (SPR) and single-cycle kinetics, we found that the rateconstants for GPIHBP11–131 and 5D2-captured LPL were fast i.e.,kon >10

7 M−1 s−1 (Fig. 2D). A parallel analysis of GPIHBP11–131/Y18F

revealed a threefold increase in Kd, a consequence of changes in bothkon and koff (Fig. 2E). This difference persisted when the rate con-stants were measured on different days with different protein prep-arations and different sensor chips (Table 2).To confirm this finding with an orthogonal method, we used

microscale thermophoresis (MST) to assess the impact of Tyr18

modifications on the capacity of synthetic GPIHBP11–33 peptidesto inhibit equilibrium binding between LPL and GPIHBP11–131

(Fig. 2F). In keeping with the SPR results (Fig. 2 D and E), theinhibitory potential was greatest for GPIHBP11–33 peptidesin which Tyr18 was sulfated or phosphorylated (IC50 0.3 μM) andlower for peptides containing unmodified Tyr18, Phe18, or Glu18

(IC50 0.6–0.7 μM).

The impact of GPIHBP1’s acidic IDR on LPL binding is bestillustrated by the fact that the LPL–GPIHBP1 encounter rateis >250-fold greater with GPIHBP11–131 (kon 2.5 × 107 M−1 s−1)than with a truncated GPIHBP134–131 (kon 0.008 × 107 M−1 s−1).This difference represents a conservative estimate, given thatmass-transport limitations is more pronouncedly slowing inter-actions driven by very fast kon in SPR analyses. Extrapolationfrom experiments conducted at higher ionic strengths suggestedthat the actual kon for the GPIHBP11–131–LPL interaction is atleast 10-fold faster (i.e., 3 × 108 M−1 s−1) at 150 mM NaCl (Fig.2G). It is likely that electrostatic forces between charged surfacesdrive this fast association rate (i.e., the acidic IDR inGPIHBP1 and the basic heparin-binding motifs in LPL) (26).Increasing ionic strength increases Kd primarily by decreasingkon, with only a minor impact on koff (Fig. 2G and Table 2); theseeffects are a hallmark of interactions controlled by electrostaticsteering (27, 28).

Protecting Against ANGPTL4-Catalyzed LPL Unfolding with GPIHBP1.We showed previously that the N-terminal acidic IDR in GPIHBP1protects LPL from spontaneous and ANGPTL4-catalyzed unfold-ing and inactivation (12, 17). To measure the relevance of theTyr18-OSO3 modification in this process, we used the samepulse-labeled hydrogen–deuterium exchange/mass spectrome-try (HDX-MS) protocol that we developed to assess LPLunfolding (12, 17). In brief, 10 μM LPL was incubated with2 μM ANGPTL41–159 for 10 min at 25 °C in protiated solventsfollowed by a 10-s pulse labeling in 70% D2O. Based on thebimodal isotope envelopes for LPL peptide 131–165, we foundthat free LPL undergoes extensive (>85%) unfolding under theseconditions (12, 17). Including 30 μM GPIHBP11–131 during theLPL/ANGPTL4 incubation reduced LPL unfolding to 8 ± 2%.Under identical conditions, the protection provided by 30 μMGPIHBP11–131/Y18F was blunted (14 ± 1% unfolding of LPL)(Fig. 3A). GPIHBP11–33 peptides alone provided much lowerdegrees of protection (12), but there was a similar trend:GPIHBP11–33 peptides that were sulfated or phosphorylatedwere more potent in protecting LPL from ANGPTL4-mediatedunfolding (Fig. 3B). Because these experiments were conductedat concentrations far exceeding the Kd for GPIHBP1–LPL in-teractions, our results suggest that O-sulfation of Tyr18 assists inprotecting LPL against ANGPTL4-catalyzed unfolding.When we measured the catalytic activity of LPL in the setting

of more physiological concentrations of LPL, the effects of theTyr18-OSO3 modification were also apparent. Incubating 15 nMLPL with 15 nM ANGPTL4 in the absence of GPIHBP1 po-tently inactivated LPL, leaving only 8 ± 3% residual LPL activity.When 15 nM GPIHBP11–131 was included in the incubation, LPLactivity was protected (38 ± 2% residual activity). When we used15 nM GPIHBP11–131/Y18F, protection of LPL was diminished(22 ± 3% residual activity) (Fig. 3C). Two factors are likely tocontribute to the reduced efficiency of GPIHBP11–131/Y18F

in protecting LPL: (i) a weaker affinity for LPL and (ii) atten-uated protection of LPL from the absence of the sulfated tyro-sine within the acidic IDR (Fig. 3A). We suspect that the lattereffect is dominant because we also observed significant differ-ences in the ability of GPIHBP11–131 and GPIHBP11–131/Y18F toprotect LPL when those proteins were added to the incubationmixture at concentrations as high as 300 nM (82 ± 2% withGPIHBP11–131 vs. only 71 ± 3% with GPIHBP11–131/Y18F).

Transitioning of LPL from an HSPG–Bound State to a Complex withGPIHBP1. GPIHBP1-mediated extraction of LPL from interstitialHSPG binding sites represents an important step in the transit ofLPL to the capillary lumen (6). Earlier confocal immunofluo-rescence microscopy studies on tissues from wild-type miceshowed that GPIHBP1 and LPL are present at the apical (lu-minal) and basolateral surfaces of capillaries, whereas the LPL inGpihbp1−/− mice never reaches the capillary lumen and insteadremains bound to HSPGs in close proximity to the surface ofcells, including both parenchymal and endothelial cells (2, 6). We

Table 1. Mass spectrometry characterization of GPIHBP1

GPIHBP1 Mass, Da Recorded mass, Da Δmass, Da

1–131 14,603.66 15,722.6 80.115,642.2 −0.4

1–131degly 14,604.64 14,684.1 79.534–131 10,579.91 11,619.2 0.31–131Y18F 14,587.66 15,627.0 0.11–33 4,039.48 4,119.41* 79.93

Protein masses were recorded on a Synapt G2 mass spectrometer (Wa-ters), and the average molecular masses were calculated by MaxEnt decon-volution. The mass difference (Δ mass) was calculated after subtracting themass of the glycan on Asn58 (1,038.96 Da). Note: both GPIHBP1 and GPIHB-PY18F contain the R38G mutation to increase protein purification yields.*The monoisotopic mass for GPIHBP11−33, released from intact GPIHBP1 bytrypsin, was calculated from the (M+4H)4+ charge state.

E6022 | www.pnas.org/cgi/doi/10.1073/pnas.1806774115 Kristensen et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

Fig. 2. Binding stoichiometry and affinity of LPL–GPIHBP1 complexes. (A–C) The binding stoichiometry between LPL and GPIHBP1 was determined by nativePAGE titrating 3 μM LPL (lane 2) with increasing amounts of GPIHBP1 ligand (range, 0.5–5.0 μM) (lanes 3–11). The relative amounts of LPL–GPIHBP1 complexes(marked by asterisks) were determined by scanning the Coomassie blue-stained bands and are superimposed as black diamonds. Data for GPIHBP11−131 andGPIHBP11−33 (both with Tyr18-OSO3) and GPIHBP11−131/Y18F are shown in A–C, respectively. (D and E) Single-cycle kinetics with SPR for the interactions be-tween LPL captured on mAb 5D2 and twofold dilutions of GPIHBP11−131 (D) or GPIHBP11−131/Y18F (E). Shown are two single-cycle runs of fiveGPIHBP1 concentrations ranging from 0.125–2 nM (red curves) or 0.25–4 nM (green curves). Data were fit to a simple bimolecular interaction (black curves)with the derived kinetic constants and the residuals shown beneath the sensorgrams. (F) Inhibitory capacity of synthetic GPIHBP11−33 peptides on the LPL–GPIHBP11−131 interaction as assessed by MST. The following peptides were measured: unmodified (orange, IC50 0.71 ± 0.20 μM); Tyr18-OSO3 (red, IC50 0.31 ±0.05 μM); Tyr18–OPO3 (green, IC50 0.30 ± 0.05 μM); Tyr18Phe (blue, IC50 0.60 ± 0.16 μM); and Tyr18Glu (purple, IC50 0.70 ± 0.08 μM). Each titration shows resultsfrom four independent replicates. Titrations with Tyr18 are significantly different from those with Tyr18-OSO3 (P < 0.01, Student’s t test). (G) Impact of ionicstrength on Kd (green squares) and kon (red circles) for the LPL–GPIHBP11−131 interaction as assessed by SPR.

Kristensen et al. PNAS | vol. 115 | no. 26 | E6023

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

speculated that GPIHBP1 would be most effective in capturingLPL if the LPL bound preferentially to the HSPGs on endo-thelial cells. To explore that hypothesis, we performed additionalconfocal microscopy studies on the heart and skeletal muscle ofGpihbp1−/− mice. Once again, we found that LPL was mis-localized within the interstitial spaces. However, we also notedthat LPL associated preferentially with endothelial cells com-pared with parenchymal cells of the heart even in the absence ofGPIHBP1 (Fig. 4 and SI Appendix, Fig. S4). This scenario wasalso evident in skeletal muscle (gastrocnemius) (Fig. 5A and SIAppendix, Fig. S5). We suspect that the preferential binding ofLPL to endothelial cells could reflect higher levels of HSPGsulfation in capillary endothelial cells. HSPG composition isknown to vary among different organs and tissues (29–31), andthis could influence LPL binding (32, 33). To pursue the ideathat different levels of HSPG sulfation could affect the tran-

sitioning of LPL to GPIHBP1, we tested the ability of differentheparin preparations (with defined length but variable sulfationpatterns) to interfere with GPIHBP1–LPL interactions. As shownin Fig. 5B, removal of N-sulfate, 2-O-sulfate, or 6-O-sulfate af-fected the capability of these defined heparin derivatives to inhibitthe GPIHBP1–LPL interaction. These studies raise the possibilitythat the LPL associated with HSPGs (e.g., glypicans, syndecans)on the surface of parenchymal cells gradually moves to higher-affinity HSPGs on capillary endothelial cells [i.e., resembling themechanism of directed diffusion established for some HSPG-tethered morphogens (34)]. Once LPL reaches the basolateralsurface of capillaries, GPIHBP1 would capture LPL and escort itacross endothelial cells to the capillary lumen.To investigate the role of GPIHBP1’s acidic IDR on the

movement of LPL to the basolateral surface of capillary endo-thelial cells, we developed an SPR sensor surface that models the

Table 2. GPIHBP1 binding kinetics to LPL

GPIHBP1 kon, 106 M−1 s−1 koff, 10

−2 s−1 Kd, nM Rmax, fmol/mm2 n

GPIHBP11-131 26.5 ± 8.2 1.26 ± 0.03 0.53 ± 0.14 1.59 ± 0.19 10GPIHBP11–131;Y18F 16.7 ± 6.4 1.81 ± 0.06 1.19 ± 0.30 1.40 ± 0.13 10GPIHBP134–131 0.087 ± 0.035 0.73 ± 0.31 97.5 ± 54.3 1.62 ± 0.06 5GPIHBP11–131 (0.30 M NaCl) 13.0 ± 0.59 5.25 ± 0.25 4.0 1GPIHBP11–131 (0.50 M NaCl) 1.36 ± 0.04 6.82 ± 0.12 52 1GPIHBP11–131 (0.75 M NaCl) 0.18 ± 0.01 5.81 ± 0.03 317 1

Kinetic rate constants (kon and koff) were derived by global fitting of sensorgrams recorded by five consecutivesingle-cycle injections of GPIHBP1 on LPL that had been captured by mAb 5D2. The values shown are means ± SDsfor single-cycle analyses performed on different chips, preparations, and days. We obtained a uniform capturelevel of LPL corresponding to 150 RU (approximately 3 fmol/mm2).

Fig. 3. Mitigation of ANGPTL4-mediated LPL unfold-ing and inactivation by GPIHBP1. (A) We measured LPLunfolding with HDX-MS by quantifying the bimodaldistribution of the isotope envelopes for the LPLpeptide 131–165 (containing Ser134 and Asp158 of thecatalytic triad). Relative amounts of unfolding of10 μM LPL incubated for 10 min at 25 °C alone (graybar) or in the presence of 2 μM ANGPTL41−159 (blackbar) are shown. Incubations with ANGPTL4 were alsoperformed with 30 μM GPIHBP11−131 (wt, red bar);30 μM GPIHBP11−131;Y18F (Y18F, blue bar), or 30 μMGPIHBP134−131 (Δacid, light gray bar). (B) Impact of30 μMGPIHBP11−33 in a similar setting with Tyr18 beingunmodified (-OH), phosphorylated (-OPO3), or sulfated(-OSO3). (C) Catalytic activity of 15 nM LPL alone(100% corresponding to 51.9 U/mL) or in the presenceof 15 nM ANGPTL41−159. These incubations were per-formed with 15, 30, 45, 150, or 300 nM GPIHBP11−131

(red bars) or GPIHBP11−131;Y18F (blue bars). Numbers ofreplicates for each experiment range from 3 to 12.Comparison of data with an unpaired t test: *P ≤ 0.05;**P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001; ns, not sig-nificant.

E6024 | www.pnas.org/cgi/doi/10.1073/pnas.1806774115 Kristensen et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

dynamic state of LPL–HSPG interactions in the subendothelialspaces and the transitioning of that pool of LPL to GPIHBP1 onthe basolateral surface of capillaries. In this model, LPL binds toa high-density surface of heparin fragments (300 fmol/mm2) onthe sensor chip. The retention of LPL on this surface is largelygoverned by mass transport limitations, with any LPL dissocia-tion event quickly followed by the binding of LPL to an adjacentunoccupied heparin. Injection of GPIHBP1 over the surfacewould allow any unbound LPL to interact with GPIHBP1. If thenewly formed LPL–GPIHBP1 complexes no longer bind to thesensor surface, they would be removed by the buffer flow. Wepredict that this in vitro system resembles the movement of LPLfrom HSPGs in the subendothelial space to GPIHBP1 on thebasolateral surface of capillaries, a binding event that promptlyleads to shuttling of the LPL–GPIHBP1 complex to the capillarylumen. Our SPR system revealed that intact GPIHBP1 efficientlyextracted LPL from the heparin-bound pool on the sensor chip(Fig. 5C). Although injections of GPIHBP134–131 led to LPL binding,those binding events were not accompanied by accelerated removalof LPL from heparin. Thus, the presence of the acidic IDR wasdecisive for a GPIHBP1-mediated LPL mobilization from heparin.Posttranslational modification with Tyr18-OSO3 improved thefunction of GPIHBP1 in this process, as GPIHBP11–131/Y18Fresulted in significantly lower levels of LPL mobilization (Fig. 5C).

Space Occupied by the Acidic IDR of GPIHBP1. Although the physi-ologic importance of GPIHBP1 is well established, the structureof GPIHBP1 has not been determined. Using small-angle X-rayscattering (SAXS), we found that GPIHBP11–33 adopts a flexi-ble, disordered, and extended conformation at pH 7.4 in 136 mMNaCl with a maximum dimension (Dmax) of 65 Å and a gyrationradius (Rg) of 19.8 Å (Fig. 6 A–C and SI Appendix, Table S1).We observed no significant difference in the scattering profilesof GPIHBP11–33 peptides containing Tyr18-OSO3 and Tyr18

-OPO3, indicating that those modifications do not induce pro-nounced shifts in the ensemble.To obtain information on the flexibility of the acidic domain in

the context of GPIHBP1, we examined both GPIHBP134–131 andGPIHBP11–131 by combining in-line size-exclusion chromatographywith SAXS (SEC-SAXS). The scattering data for GPIHBP134–131(Fig. 6E) provided a suboptimal match to our model for the LUdomain in GPIHBP1 (17) (CRYSOL; χ2 4.6). By adjusting for theflexibility of the small biantennary glycan on Asn58 [with theAllosMod-FoXS server (35)], a moderately improved fit wasobtained (χ2 3.9). Guided by earlier HDX-MS analyses (17), we

allowed flexibility in certain regions of the peptide backbone of theGPIHBP134–131 model (Fig. 6D). Applying the ensemble optimi-zation method (EOM) for this condition led to a substantial im-provement in the fit (χ2 2.4). Superimposing all ensemblesselected during 10 separate rounds of EOM revealed that the C-terminal region of GPIHBP1 (residues 113–131) projected awayfrom the central β-sheet of GPIHBP134–131 (Fig. 6E and SI Ap-pendix, Table S3). Of note, this orientation is recapitulated byindependent EOM analysis with intact GPIHBP11–131 (Fig. 6G).These data imply that GPIHBP1 projects its LPL-binding in-terface [defined by HDX-MS and site-directed mutagenesis (17,36)] away from the cell surface.Fitting the scattering data for full-length GPIHBP11–131 with the

EOM revealed that the acidic domain occupies a large mushroom-shaped space with a diameter of 112 Å (Fig. 6 F and G). Thisprediction aligns well with the scattering data for GPIHBP11–33

peptides, which exhibit a Dmax of 65 Å. From a functional per-spective, we propose that the large space occupied by the acidicIDR is important for recruiting LPL from the subendothelialHSPG binding sites via long-range electrostatic steering, facilitatinga subsequent high-affinity interaction with GPIHBP’s LU domain.

Reactivity of GPIHBP1 Autoantibodies. Due to their low sequencecomplexity and paucity of hydrophobic residues, IDRs are gen-erally poor ligands for MHC class II and are likely to escapeadaptive immune responses (37, 38). In keeping with our SAXSdata, the first 35 residues of human GPIHBP1 are predicted tobe highly disordered, translating into a low predicted affinity forMHC class II (Fig. 7A). Despite these considerations, a mousemonoclonal antibody against human GPIHBP1 (mAb RF4) wasproposed to bind a linear epitope within GPIHBP1’s acidic do-main (based on the observation that the antibody bound to full-length GPIHBP1 but not to a mutant lacking the acidic domain)(39). Because IDRs such as GPIHBP1’s acidic domain are notexpected to elicit a strong immune response, we revisited thelocation of the mAb RF4 epitope. With SPR, we compared thekinetics of mAb RF4 binding to synthetic peptides derived fromthe acidic IDR of GPIHBP1. Our studies mapped the epitopefor mAb RF4 to residues 27–44 within GPIHBP1, with Arg33 andLeu34 being the hot-spot residues, and those residues are notconserved in mouse GPIHBP1 (Fig. 7 B and C and SI Appendix,Fig. S6). Thus, the epitope for RF4 is located downstream fromthe acidic IDR, in the region just upstream of the folded LUdomain (arrow in Fig. 7A).We recently reported that some patients develop chylomi-

cronemia as a result of GPIHBP1 autoantibodies that block LPLbinding (GPIHBP1 autoantibody syndrome) (15, 16). To assessthe involvement of the IDR in this disease, we captured anti-bodies from an affected patient on a Protein G sensor chip. In-jections of GPIHBP1 revealed that 2.2 ± 0.3% of the patient’sIgGs bound to GPIHBP1 (Fig. 7D). In keeping with that result,affinity purification of the total IgG fraction on a humanGPIHBP1–Sepharose column resulted in a 1.2% yield of antibodiesrecognizing GPIHBP1. When analyzed by SPR, the immunopuri-fied autoantibodies bound with high affinity to GPIHBP134–131,whereas there was no binding to GPIHBP11–45 (Fig. 7E). The sameproperties (binding to GPIHBP134–131 but not to GPIHBP11–45)were observed with the IgGs purified from the plasma of a secondpatient with the GPIHBP1 autoantibody syndrome (SI Appendix,Fig. S7). These findings support the idea that GPIHBP1’s acidicdomain is not immunogenic and that autoantibodies against theacidic domain are not relevant to the pathogenesis of this autoim-mune chylomicronemia syndrome.

DiscussionGPIHBP1 plays at least three distinct roles in the intravascularprocessing of triglyceride-rich lipoproteins. First, GPIHBP1 bindsLPL in the subendothelial spaces and shuttles it to its site of action inthe capillary lumen (6). Second, GPIHBP1–LPL complexes incapillaries mediate the margination of chylomicrons along capillaries(5). Third, GPIHBP1 protects LPL from two physiologic inhibitor

Fig. 4. Localization of LPL in the mouse heart. The distribution of LPL in thehearts of wild-type and Gpihbp1−/− (KO) mice was assessed by immunohis-tochemistry with antibodies against LPL (red), CD31 (magenta), or β-dystroglycan(green). Nuclei were stained with DAPI (blue). Shown are confocal fluorescencemicroscopy images of capillary endothelial cells containing an endothelial cellnucleus (which makes it possible to visualize the basolateral and apical mem-branes). In the wild-type mouse heart, LPL was associated almost exclusively withcapillary endothelial cells (arrow). In the heart of a Gpihbp1-deficient mouse, theLPL is mislocalized within the interstitial spaces surrounding both myocytes andcapillary endothelial cells, but LPL appeared to bind in close proximity to capillaryendothelial cells (arrows). (Scale bars, 5 μm.)

Kristensen et al. PNAS | vol. 115 | no. 26 | E6025

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

proteins, ANGPTL3 and ANGPTL4 (12, 17). The acidic IDR withinGPIHBP1, the focus of the current studies, contributes to the reg-ulation of plasma triglyceride metabolism at all three levels.We now demonstrate that electrostatic steering boosts the en-

counter rate between GPIHBP1 and LPL, increasing the associ-ation rate constant by >250-fold, a finding that is likely relevant tothe interactions between GPIHBP1 and LPL in the subendothelialspaces. The acidic IDR in GPIHBP1 is the key factor in drivingthe accelerated kinetics of LPL binding and almost certainly doesso by transient interactions with LPL’s basic heparin-binding re-gions. Our SAXS analyses revealed that GPIHBP1’s acidic IDRspans a relatively large conformational space (112 Å in diameter),which likely augments its capacity to capture LPL within thesubendothelial spaces. We also documented the existence of ahitherto unrecognized tyrosyl-O-sulfation near the center ofGPIHBP1’s IDR and showed that this modification slightly in-creases the affinity of GPIHBP1–LPL interactions. We expect thatthe combined impact of these biochemical properties rendersGPIHBP1 highly adept for extracting LPL from its HSPG-tethered reservoir in the subendothelial spaces. The impor-tance of GPIHBP1’s IDR in recruiting LPL to GPIHBP1 issupported by the observation that full-length GPIHBP1 removesLPL from heparin on SPR sensor chips, whereas GPIHBP134–131

does not. We speculate that GPIHBP1’s acidic IDR is crucial forextracting LPL from the subendothelial spaces, explaining whythis domain has been so strongly conserved during mammalianevolution.With a view to LPL partitioning in the subendothelial spaces of

myocytes, we observed by confocal immunofluorescence micros-copy that LPL accumulates in close proximity to the capillary en-dothelial cells of heart and skeletal muscle even when GPIHBP1 isabsent. This finding implies directional movement of interstitialLPL to capillary endothelial cells. The mechanism underlying di-rectional movement to endothelial cells is unknown, but we suspectthat the mechanism involves the same general factors shaping theconcentration gradients of morphogens during tissue differentiationand embryogenesis (34, 40, 41). In the latter process, the densityand sulfation profile of HSPGs are important factors, along withextracellular processing of HSPGs by secreted endosulfatases (42).In triglyceride metabolism, HSPGs were for many decades con-sidered the primary binding site for LPL within the capillary lumen(43), and there was little discussion of HSPGs in the subendothelialspaces. Subsequent studies showed that GPIHBP1, not HSPGs, isthe binding site for LPL inside capillaries (4–6). Interestingly, thosesame studies cast a spotlight on the relevance of subendothelialHSPGs in LPL trafficking. In GPIHBP1-deficient mice, the newlysecreted LPL is not removed from the subendothelial spaces bylymph drainage but remains bound, via a transient interaction withHSPGs. The current studies add to our understanding of LPL in-teractions within the interstitium. In GPIHBP1-deficient mice, LPLassociates preferentially with capillaries, implying that LPL movesin a directional fashion from the HSPGs surrounding paren-chymal cells to HSPGs near capillary endothelial cells. CollagenXVIII and perlecan are the two major HSPGs found in vascularbasement membranes. We would not be surprised if collagenXVIII plays a role in the directional movement of LPL towardcapillaries, because Col18a1−/− mice have increased plasmatriglyceride levels along with reduced transport of LPL to thecapillary lumen (44).The LPL–GPIHBP1 complex in the capillary lumen is re-

quired for margination of TRLs along capillaries. This dockingprocess occurs despite vascular shear stress and resembles the

Fig. 5. Movement of LPL between HSPGs and GPIHBP1. (A) The distributionof LPL in gastrocnemius muscle of wild-type and Gpihbp1−/− (KO) mice wasassessed by confocal immunofluorescence microscopy with antibodiesagainst LPL (green) and CD31 (red). Nuclei were stained with DAPI (blue). Inthe wild-type mouse LPL was associated almost exclusively with capillaryendothelial cells. In the Gpihbp1-deficient mouse the LPL is mislocalizedwithin the interstitial spaces surrounding both myocytes and capillary en-dothelial cells, but LPL appeared to bind preferentially to capillary endo-thelial cells. (Scale bars, 20 μm.) (B) Competition of the GPIHBP1–LPLinteraction by defined heparin fragments (dp10) by MST. Heparin (red curve;IC50 <14 nM), desulfated on the C2 oxygen of iduronate (green curve; IC50

59 ± 5 nM), desulfated on the C6 oxygen of glucosamine (blue curve; IC50

100 ± 10 nM), and desulfated on the C2 amine of glucosamine (orangecurve; IC50 980 ± 14 nM). (C) Mobilization of LPL from a high-density heparinsurface by the injection of 200 nM of various GPIHBP1 derivatives. After

1,000-s exposures of GPIHBP1 at a flowrate of 20 μL/min (gray line), differentlevels of LPL remained on the heparin surface: 23.1% by GPIHBP11−131 (bluecurve); 27.3%by GPIHBP11−131; Y18F (red curve); and 97.0% by GPIHBP134−131 (blackcurve). Although the difference between GPIHBP11−131 and GPIHBP11−131; Y18F

was modest (2.5 ± 1.6%), it was highly significant when comparing10 consecutive runs with a paired t test (P < 0.001). Buffer control is shownby the green curve.

E6026 | www.pnas.org/cgi/doi/10.1073/pnas.1806774115 Kristensen et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

initial phase of leukocyte extravasation in the setting of in-flammation. Here transient interactions between endothelial cellP-selectin and the leukocyte mucin PSGL-1 cause rolling ofleukocytes along the vascular endothelium (45). Like GPIHBP1,PSGL-1 contains an N-terminal IDR that is relatively acidic withthree sulfated tyrosines. Interestingly, the affinity of PSGL-1 for P-selectin increases two- to fivefold by cooperative binding via itssulfated tyrosine residues and an adjacent O-linked glycan (46,47). This cooperativity is functionally important and translates intomore efficient leukocyte rolling (45). Whether any cooperativityexists between marginated TRLs and the LPL–GPIHBP1 complexon endothelial cells, and whether GPIHBP1’s IDR or its tyrosyl-O-sulfate plays a role in this process, is unclear. However, it isnoteworthy that several of the apolipoproteins on the surface ofTRLs have heparin-binding motifs, and it is quite conceivable thatthe fast binding kinetics that characterizes interactions betweenGPIHBP1’s IDR and LPL also applies to interactions with TRLapolipoproteins. It is easy to imagine a scenario in which theformation of a ternary GPIHBP1–LPL–TRL complex constitutesthe functional unit for hydrolyzing the triglycerides within TRLs.Circumstantial evidence supports this possibility (48–50).Finally, we showed that the sulfate moiety on GPIHBP1 Tyr18

improves the capacity of GPIHBP1 to protect LPL againstANGPTL4-mediated unfolding, thereby improving the protectionand longevity of the catalytic activity of GPIHBP1-bound LPLwithin capillaries. Such functional importance suggests that thesulfated tyrosine in GPIHBP1’s IDR could engage LPL in aspecific, albeit transient, interaction and thus act to preservethe structural integrity of LPL. Evolutionarily conserved motifs

within IDRs are often functionally important (51), and the ty-rosine sulfation in GPIHBP1 may represent one more exampleof this correlation.Posttranslational modification with tyrosine sulfate occurs in the

trans-Golgi network where two tyrosylprotein sulfotransferases,TPST1 and TPST2, catalyze the sulfate transfer from an activateddonor. This molecule (3′-phosphoadenosine-5′-phosphosulfate) istransported to the trans-Golgi from the cytosol by the solute carrierfamily 35 member B2 (SLC35B2). It would therefore be interestingto test if loss-of-function mutations in TPST or SLC35B2 are as-sociated with increased plasma triglyceride levels. A genome-wideCRISPER screen identified TPST2 and SLC35B2 as essential hostdependency factors for the entry of HIV, and this was causallyrelated to the tyrosine sulfation of CCR5 (52).The current studies add substantially to our understanding of

the biochemistry and physiology of intravascular triglyceridemetabolism, but key mechanistic issues remain unsolved. Forexample, how does ANGPTL4 catalyze an ATP-independentprotein unfolding regulating LPL activity? Also, how doesGPIHBP1 antagonize this unfolding? The answers are unknown,but our studies would suggest that GPIHBP1’s IDR and its sul-fated tyrosine play central roles in this process.

Materials and MethodsPurified Proteins and Synthetic Peptides. Recombinant and secreted versionsof human GPIHBP11–131/R38G and murine GPIHBP11–178 were produced in Dro-sophila S2-cells as fusion proteins with an N-terminal human uPAR D3 tag (53)and purified as described (17). Bovine LPL (bLPL) was purified from fresh bovinemilk (54). The coiled-coil domain of ANGPTL4 (residues 1–159 with an N-terminal

Fig. 6. SAXS analyses of GPIHBP1. (A) Concentration-normalized scattering profiles of GPIHBP11−33 with Tyr18–OH (2.6 mg/mL, black curve); Tyr18-OSO3 (3.8 mg/mL,blue curve); or Tyr18–OPO3 (3.6 mg/mL, green curve). (B) Corresponding pair–distance distribution functions [p(r)]. (C) Kratky plot illustrating the high flexibility anddisorder of these peptides, which is best described by an ideal random-walk chain structure (58). (D) Heat-map representation of the flexibility in GPIHBP11−131

determined by HDX-MS (17), high exchange (red) to low exchange (blue). Black lines highlight sequences allowed to be flexible during EOM simulations of the SAXSdata. (E) SEC-SAXS scattering data for a truncated GPIHBP1 lacking the acidic domain (GPIHBP134−131, black circles) along with the scattering profile of a rigidGPIHBP1 homology model (17) (CRYSOL; blue curve, χ2 4.6), a model allowing glycosylation flexibility (Allosmod; red curve, χ2 3.9), and a model allowing definedpeptide flexibility (EOM; green curve, χ2 2.4). The Inset shows models selected by 10 separate EOM analyses; a cartoon representation shows the nonvariable part ofGPIHBP134−131, and dots (residues 34–42 and 74–83) or spheres (118–131) show the variable parts. The position of Trp89, which is important for LPL binding, is shownin a stick representation. (F) SEC-SAXS scattering data for full-length GPIHBP11−131 with fits to a rigid homology model (CRYSOL; blue curve, χ2 10.6) and 10 EOManalyses (green curve, χ2 1.6). (G) Conformational ensembles selected for our current model of GPIHBP11−131 by the 10 separate EOMs. The illustration was preparedby PyMOL (Schrödinger) using the same settings as in E except that the spheres show residues 1–42 and dots show residues 74–83 and 118–131.

Kristensen et al. PNAS | vol. 115 | no. 26 | E6027

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

methionine and a C-terminal 6× His-tag) was produced in Escherichia coli (55). Amouse monoclonal antibody against sulfotyrosine (mAb 1C-A2) was from MerckMillipore. Defined heparan sulfate oligomers were from Iduron and are listed inSI Appendix, Table S3.

Synthetic peptides representing various truncations andmodifications of theN-terminal acidic domain of GPIHBP1 were obtained at a purity of >95% fromTAG-Copenhagen A/S and are listed in SI Appendix, Table S3. Peptides withtyrosine sulfate modifications were prepared in-house by standard solid-phasepeptide Fmoc synthesis with orthogonal protection of sulfotyrosine residueswith a neopentyl-protected cassette [Fmoc–Tyr(OSO3nP)–OH] as described (56).Quantitative derivatization of Tyr18 in GPIHBP11–33 with tyrosyl-O-sulfate wasverified by 1H-NMR (SI Appendix, Fig. S8). Due to the lack (or paucity) of ar-omatic amino acids, peptide concentrations were determined by absorbanceat 214 nm using calculated molar extinction coefficients (57).

Kinetics of LPL–GPIHBP1 Binding Assessed by SPR. Reaction kinetics betweenLPL and GPIHBP1 derivatives were measured with a Biacore T200 instrumentusing a modified version of a previous protocol (17). In brief, we immobilizedan anti-LPL monoclonal antibody (5D2) on a CM4 chip and then captured150 nM bLPL in 10 mM Hepes (pH 7.4), 300 mM NaCl, 4 mM CaCl2, 10% (vol/vol)glycerol, 0.05% (vol/vol) surfactant P20, 1 mg/mL defatted BSA, 0.1 mg/mL car-boxylmethyl dextran, and 0.05% (wt/vol) NaN3. This protocol kept the LPL cat-alytically active and led to capture densities of 150 resonance units (RUs)(approximately 3 fmol LPL/mm2). We used single-cycle kinetics to determine theinteraction between the captured LPL and the various GPIHBP1 derivatives in10 mM Hepes (pH 7.4), 150 mM NaCl, 4 mM CaCl2, 0.05% (vol/vol) P20, 0.2 mg/mLdefatted BSA, and 0.05% (wt/vol) NaN3. This protocol included five con-secutive injections of twofold dilutions of GPIHBP1 (ranging from 0.125–2 nM and/or 0.25–4 nM) at a flowrate of 50 μL/min at 20 °C. At the end ofeach cycle, two 10-μL injections of 20-mM H3PO4 regenerated the chip. Therecorded sensorgrams were double-buffer referenced, and the bindingrate constants were calculated by fitting the data to a simple bimolecular

interaction model with the mathematical model developed for single-cyclekinetics (T200 Evaluation Software 3.0; GE Healthcare).

GPIHBP1-Mediated Mobilization of LPL from a High-Density HSPG Surface. Tocreate a surrogate high-density HSPG surface and thusmimic the conditions inthe subendothelial spaces, we captured well-defined biotinylated heparindp4 fragments (SI Appendix, Table S3) on a streptavidin-coupled CM5 sensorchip. The reference surface contained the nonsulfated fragment M09 S00,and the active surface contained M09 S08a with N- and O-6-sulfations(Iduron). Only the sulfated oligosaccharide bound LPL, and this yieldedonly weak and transient interactions. We chose to use very high surfacedensities (300 fmol/mm2) of the two heparin fragments to enable the sur-face confinement of LPL via a dominating mass transport limitation. In eachcycle, 100 nM LPL was loaded for 50 s at a flow rate of 50 μL/min in 10 mMHepes (pH 7.4), 150 mMNaCl, 4 mM CaCl2, 10% (vol/vol) glycerol, 0.05% (vol/vol)surfactant P20, 1 mg/mL defatted BSA, 1 μM GPIHBP11–33:Tyr-OH, and 0.05%(wt/vol) NaN3. The synthetic GPIHBP1 peptide stabilized the LPL-loadingsample during the SPR experiment, ensuring a uniform capture level. Thisprocedure resulted in capture levels of 13 fmol/mm2 LPL. The ability ofGPIHBP1 to bind and extract LPL from this reservoir was tested by injecting200 nM of GPIHBP1 for 1,000 s at 20 μL/min in 10 mM Hepes (pH 7.4),150 mM NaCl, 4 mM CaCl2, 10% (vol/vol) glycerol, 0.05% (vol/vol) surfac-tant P20, 0.2 mg/mL defatted BSA, and 0.05% (wt/vol) NaN3. At the end ofeach cycle, two consecutive injections of 10 μL 1-M NaCl and 3-M guani-dinium chloride regenerated the chip.

Ethical Considerations. All mouse studies were approved by University ofCalifornia, Los Angeles (UCLA)’s animal research committee. Human plasmasamples were received at UCLA without identifiers and were thereforedeemed exempt from human use approval by UCLA.

ACKNOWLEDGMENTS. We thank Gry Rasmussen and Gitte W. Haxholm forexpert technical assistance, John R. Couchman (Biotech Research and

Fig. 7. Reactivity of GPIHBP1 autoantibodies purifiedfrom a subject with the GPIHBP1 autoantibody syn-drome. (A) A comparison of the predicted IDRs andreactivity with MHC class II for GPIHBP1. Disorder pre-diction was assessed by IUPred (59), and reactivity witha MHC class II receptor (HLA-DRB1101 allele) was pre-dicted by NetMHCIIpan version 3.1 (60) using a 15-mersequence window. The cyan boxes show the positionsof predicted secondary structure elements (β-strands).(B and C) Definition of the domain reactivity of amonoclonal antibody against humanGPIHBP1 (mAb RF4)by single-cycle kinetics of twofold dilutions (2–32 nM) ofGPIHBP11−131 (red curve in B), GPIHBP11−45 (green curvesin B and C), GPIHBP134−131 (blue curve in B), GPIHBP127−44

(cyan curve in C), and GPIHBP127−44/R33M (blue curve in C).(D) Single-cycle kinetics of 2–32 nM GPIHBP134−131 bind-ing to Protein G–captured total IgG (2 μg/mL) isolatedeither from a patient with GPIHBP1 autoantibody syn-drome [patient 102 (15), red curve] or from a healthynormolipidemic control subject (blue curve). The Insetshows the level of Protein G–captured immunoglobulins,and the box represents the GPIHBP1 binding segment,which is enlarged in the main figure. Comparisons ofcapture levels and the calculated binding capacities forGPIHBP1 reveal that 2.2 ± 0.3% (n = 9) of the total IgGfraction binds GPIHBP134−131. (E) Binding profiles foraffinity-purified GPIHBP1 autoantibodies to 2–32 nMGPIHBP134−131 (red curve) or to GPIHBP11−45 (blue curve).These studies showed that 78 ± 7% (n = 4) of theaffinity-purified IgG binds to GPIHBP134−131 (red curve),whereas none of the autoantibodies binds to GPIHBP1’sacidic IDR. Thin black lines in B–E show the kinetic fit ofthe data to a 1:1 binding model.

E6028 | www.pnas.org/cgi/doi/10.1073/pnas.1806774115 Kristensen et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1

Innovation Centre, University of Copenhagen) for critical comments, andHenrik Gårdsvoll for preparing the expression vector for GPIHBP1Y18F. This workwas supported by Leducq Transatlantic Network Grant 12CVD04, Lundbeck

Foundation Grant R230-2016-2930, NOVO Nordisk Foundation GrantNNF17OC0026868, Swedish Research Council Grant 2015-02942, and Na-tional Heart, Lung, and Blood Institute Grants HL090553 and HL087228.

1. Brahm AJ, Hegele RA (2015) Chylomicronaemia–Current diagnosis and future thera-pies. Nat Rev Endocrinol 11:352–362.

2. Fong LG, et al. (2016) GPIHBP1 and plasma triglyceride metabolism. Trends EndocrinolMetab 27:455–469.

3. Olivecrona G (2016) Role of lipoprotein lipase in lipid metabolism. Curr Opin Lipidol27:233–241.

4. Beigneux AP, et al. (2007) Glycosylphosphatidylinositol-anchored high-densitylipoprotein-binding protein 1 plays a critical role in the lipolytic processing of chylo-microns. Cell Metab 5:279–291.

5. Goulbourne CN, et al. (2014) The GPIHBP1-LPL complex is responsible for the mar-gination of triglyceride-rich lipoproteins in capillaries. Cell Metab 19:849–860.

6. Davies BSJ, et al. (2010) GPIHBP1 is responsible for the entry of lipoprotein lipase intocapillaries. Cell Metab 12:42–52.

7. He C, et al. (2018) NanoSIMS analysis of intravascular lipolysis and lipid movementacross capillaries and into cardiomyocytes. Cell Metab 27:1055–1066.e3.

8. Dijk W, et al. (2015) ANGPTL4 mediates shuttling of lipid fuel to brown adipose tissueduring sustained cold exposure. eLife 4:e08428.

9. Gusarova V, et al. (2015) ANGPTL3 blockade with a human monoclonal antibodyreduces plasma lipids in dyslipidemic mice and monkeys. J Lipid Res 56:1308–1317.

10. Wang Y, et al. (2013) Mice lacking ANGPTL8 (Betatrophin) manifest disrupted tri-glyceride metabolism without impaired glucose homeostasis. Proc Natl Acad Sci USA110:16109–16114.

11. Sukonina V, Lookene A, Olivecrona T, Olivecrona G (2006) Angiopoietin-like protein4 converts lipoprotein lipase to inactive monomers and modulates lipase activity inadipose tissue. Proc Natl Acad Sci USA 103:17450–17455.

12. Mysling S, et al. (2016) The angiopoietin-like protein ANGPTL4 catalyzes unfolding ofthe hydrolase domain in lipoprotein lipase and the endothelial membrane proteinGPIHBP1 counteracts this unfolding. eLife 5:e20958.

13. Sonnenburg WK, et al. (2009) GPIHBP1 stabilizes lipoprotein lipase and prevents itsinhibition by angiopoietin-like 3 and angiopoietin-like 4. J Lipid Res 50:2421–2429.

14. Allan CM, et al. (2017) Mutating a conserved cysteine in GPIHBP1 reduces amounts ofGPIHBP1 in capillaries and abolishes LPL binding. J Lipid Res 58:1453–1461.

15. Beigneux AP, et al. (2017) Autoantibodies against GPIHBP1 as a cause of hyper-triglyceridemia. N Engl J Med 376:1647–1658.

16. Hu X, et al. (2017) GPIHBP1 autoantibodies in a patient with unexplained chylomi-cronemia. J Clin Lipidol 11:964–971.

17. Mysling S, et al. (2016) The acidic domain of the endothelial membrane proteinGPIHBP1 stabilizes lipoprotein lipase activity by preventing unfolding of its catalyticdomain. eLife 5:e12095.

18. Babu MM (2016) The contribution of intrinsically disordered regions to proteinfunction, cellular complexity, and human disease. Biochem Soc Trans 44:1185–1200.

19. Kulkarni P, et al. (2017) Phosphorylation-induced conformational dynamics in an in-trinsically disordered protein and potential role in phenotypic heterogeneity. ProcNatl Acad Sci USA 114:E2644–E2653.

20. Bah A, et al. (2015) Folding of an intrinsically disordered protein by phosphorylationas a regulatory switch. Nature 519:106–109.

21. Mabashi-Asazuma H, Kuo CW, Khoo KH, Jarvis DL (2015) Modifying an insect cell N-glycan processing pathway using CRISPR-Cas technology. ACS Chem Biol 10:2199–2208.

22. Gårdsvoll H, Werner F, Søndergaard L, Danø K, Ploug M (2004) Characterization oflow-glycosylated forms of soluble human urokinase receptor expressed in DrosophilaSchneider 2 cells after deletion of glycosylation-sites. Protein Expr Purif 34:284–295.

23. Kehoe JW, et al. (2006) Using phage display to select antibodies recognizing post-translational modifications independently of sequence context. Mol Cell Proteomics5:2350–2363.

24. Monigatti F, Gasteiger E, Bairoch A, Jung E (2002) The sulfinator: Predicting tyrosinesulfation sites in protein sequences. Bioinformatics 18:769–770.

25. Pan Z, et al. (2014) Systematic analysis of the in situ crosstalk of tyrosine modificationsreveals no additional natural selection on multiply modified residues. Sci Rep 4:7331.

26. Zhou HX, Bates PA (2013) Modeling protein association mechanisms and kinetics. CurrOpin Struct Biol 23:887–893.

27. Schreiber G, Haran G, Zhou HX (2009) Fundamental aspects of protein-protein asso-ciation kinetics. Chem Rev 109:839–860.

28. Borgia A, et al. (2018) Extreme disorder in an ultrahigh-affinity protein complex.Nature 555:61–66.

29. Iozzo RV, Schaefer L (2015) Proteoglycan form and function: A comprehensive no-menclature of proteoglycans. Matrix Biol 42:11–55.

30. Dallinga MG, Dallinga-Thie GM (2016) Role of sulfatase 2 in lipoprotein metabolismand angiogenesis. Curr Opin Lipidol 27:181–186.

31. Ledin J, et al. (2004) Heparan sulfate structure in mice with genetically modifiedheparan sulfate production. J Biol Chem 279:42732–42741.

32. Lookene A, Chevreuil O, Ostergaard P, Olivecrona G (1996) Interaction of lipoproteinlipase with heparin fragments and with heparan sulfate: Stoichiometry, stabilization,and kinetics. Biochemistry 35:12155–12163.

33. Spillmann D, Lookene A, Olivecrona G (2006) Isolation and characterization of lowsulfated heparan sulfate sequences with affinity for lipoprotein lipase. J Biol Chem281:23405–23413.

34. Duchesne L, et al. (2012) Transport of fibroblast growth factor 2 in the pericellularmatrix is controlled by the spatial distribution of its binding sites in heparan sulfate.PLoS Biol 10:e1001361.

35. Guttman M, Weinkam P, Sali A, Lee KK (2013) All-atom ensemble modeling to ana-lyze small-angle X-ray scattering of glycosylated proteins. Structure 21:321–331.

36. Beigneux AP, et al. (2011) Assessing the role of the glycosylphosphatidylinositol-anchored high density lipoprotein-binding protein 1 (GPIHBP1) three-finger domainin binding lipoprotein lipase. J Biol Chem 286:19735–19743.

37. Miti�c NS, Pavlovi�c MD, Jandrli�c DR (2014) Epitope distribution in ordered and disor-dered protein regions–Part A. T-cell epitope frequency, affinity and hydropathy.J Immunol Methods 406:83–103.

38. Guy AJ, et al. (2015) Insights into the immunological properties of intrinsically dis-ordered malaria proteins using proteome scale predictions. PLoS One 10:e0141729.

39. Hu X, et al. (2017) Monoclonal antibodies that bind to the Ly6 domain ofGPIHBP1 abolish the binding of LPL. J Lipid Res 58:208–215.

40. Balasubramanian R, Zhang X (2016) Mechanisms of FGF gradient formation duringembryogenesis. Semin Cell Dev Biol 53:94–100.

41. Yan D, Lin X (2009) Shaping morphogen gradients by proteoglycans. Cold Spring HarbPerspect Biol 1:a002493.

42. Kleinschmit A, et al. (2010) Drosophila heparan sulfate 6-O endosulfatase regulatesWingless morphogen gradient formation. Dev Biol 345:204–214.

43. Merkel M, Eckel RH, Goldberg IJ (2002) Lipoprotein lipase: Genetics, lipid uptake, andregulation. J Lipid Res 43:1997–2006.

44. Bishop JR, et al. (2010) Deletion of the basement membrane heparan sulfate pro-teoglycan type XVIII collagen causes hypertriglyceridemia in mice and humans. PLoSOne 5:e13919.

45. Zarbock A, Ley K, McEver RP, Hidalgo A (2011) Leukocyte ligands for endothelialselectins: Specialized glycoconjugates that mediate rolling and signaling under flow.Blood 118:6743–6751.

46. Somers WS, Tang J, Shaw GD, Camphausen RT (2000) Insights into the molecular basisof leukocyte tethering and rolling revealed by structures of P- and E-selectin bound toSLe(X) and PSGL-1. Cell 103:467–479.

47. Sako D, et al. (1995) A sulfated peptide segment at the amino terminus of PSGL-1 iscritical for P-selectin binding. Cell 83:323–331.

48. Shu X, et al. (2010) Intravenous injection of apolipoprotein A-V reconstituted high-density lipoprotein decreases hypertriglyceridemia in apoav-/- mice and requiresglycosylphosphatidylinositol-anchored high-density lipoprotein-binding protein 1.Arterioscler Thromb Vasc Biol 30:2504–2509.

49. Gin P, et al. (2008) The acidic domain of GPIHBP1 is important for the binding of li-poprotein lipase and chylomicrons. J Biol Chem 283:29554–29562.

50. Gin P, et al. (2011) Binding preferences for GPIHBP1, a glycosylphosphatidylinositol-anchored protein of capillary endothelial cells. Arterioscler Thromb Vasc Biol 31:176–182.

51. Nguyen Ba AN, et al. (2012) Proteome-wide discovery of evolutionary conserved se-quences in disordered regions. Sci Signal 5:rs1.

52. Park RJ, et al. (2017) A genome-wide CRISPR screen identifies a restricted set of HIVhost dependency factors. Nat Genet 49:193–203.

53. Gårdsvoll H, Hansen LV, Jørgensen TJ, Ploug M (2007) A new tagging system forproduction of recombinant proteins in Drosophila S2 cells using the third domain ofthe urokinase receptor. Protein Expr Purif 52:384–394.

54. Bengtsson-Olivecrona G, Olivecrona T (1991) Phospholipase activity of milk lipopro-tein lipase. Methods Enzymol 197:345–356.

55. Robal T, Larsson M, Martin M, Olivecrona G, Lookene A (2012) Fatty acids bind tightlyto the N-terminal domain of angiopoietin-like protein 4 and modulate its interactionwith lipoprotein lipase. J Biol Chem 287:29739–29752.

56. Stone MJ, Payne RJ (2015) Homogeneous sulfopeptides and sulfoproteins: Syntheticapproaches and applications to characterize the effects of tyrosine sulfation onbiochemical function. Acc Chem Res 48:2251–2261.

57. Kuipers BJ, Gruppen H (2007) Prediction of molar extinction coefficients of proteinsand peptides using UV absorption of the constituent amino acids at 214 nm to enablequantitative reverse phase high-performance liquid chromatography-mass spec-trometry analysis. J Agric Food Chem 55:5445–5451.

58. Kohn JE, et al. (2004) Random-coil behavior and the dimensions of chemically un-folded proteins. Proc Natl Acad Sci USA 101:12491–12496.

59. Dosztányi Z, Csizmok V, Tompa P, Simon I (2005) IUPred: Web server for the pre-diction of intrinsically unstructured regions of proteins based on estimated energycontent. Bioinformatics 21:3433–3434.

60. Andreatta M, et al. (2015) Accurate pan-specific prediction of peptide-MHC class IIbinding affinity with improved binding core identification. Immunogenetics 67:641–650.

Kristensen et al. PNAS | vol. 115 | no. 26 | E6029

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 14

, 202

1